High Resolution Optical Imaging Techniques for Rapid Assessment of Breast Cancer
Dobbs, Jessica Lupinacci
Doctor of Philosophy
Breast cancer is the most prevalent and deadly cancer among women worldwide. The current standard for breast lesion diagnosis is histologic assessment with hematoxylin and eosin (H&E) staining. Histology has high diagnostic accuracy, but requires extensive time and resources to perform. The objective of this work was to improve diagnosis of early breast cancers by developing approaches to rapidly image and characterize neoplastic tissue and the tumor microenvironment in high resolution optical images. Confocal fluorescence microscopy can image optical sections of tissue without the need for extensive tissue processing. Three studies were performed to evaluate if confocal microscopy images contain sufficient information to identify neoplasia in breast tissue. In a 31 patient study, five pathologists identified neoplasia with high accuracy in confocal and histologic images. In another study, an expert pathologist estimated tumor cellularity in core biopsies with moderate agreement between confocal and histologic images. In a third study, an expert pathologist assigned diagnoses and grades to neoplastic tissue in confocal and histologic images. Limitations of these studies include recruitment of patients at a single center and data assessment by a single reader in two of three studies. Visual assessment for cancer diagnosis is limited by the potential for inter- and intra-observer error. Using a computerized algorithm to segment and quantify architectural features of breast ducts and nuclei, a decision-tree model was developed that classified confocal images of breast tissue sites as neoplastic or non-neoplastic with an overall accuracy of 90%. Another computerized algorithm was developed to segment adipocytes in confocal images and results showed significant differences in phenotypic properties of adipocytes adjacent to neoplastic and non-neoplastic tissue. High resolution microendoscopy (HRME) can be used to rapidly acquire images at a lower cost than confocal microscopy. In a study evaluating HRME and two approaches to improve image contrast, results demonstrated that HRME with structured illumination yields images with high contrast relative to HRME with standard illumination. The unique contribution of these results is the characterization of qualitative and quantitative criteria to evaluate breast tissue and classify neoplasia in optical images, although recognition of invasive lobular carcinoma was limited. The criteria developed in this research may be applied to further development of techniques for objective classification and diagnosis of breast cancer in optical images.
optical imaging; breast cancer